How to make better bots (without any hard programming)

It's seems like the the bots on chess.com are just the same program set to different difficulty leves and with different dialog. Obviously it would be take an insane amount of work to actually tailor each individual bot to each individual playing style. Or would it?

Right now you can use the openings explorer and see what each of the persons these bots are based on usually do, why not just use that and a random number generator to chose the next move? That way the bots will play the openings the actual people usually play and react in the way they usually react, since you will in essence just be replaying their old games. When you reach a position the actual person haven't played that much (for example less than 10 times) then the engine can take over from there. Considering the data needed for this is something chess.com already collects and is already accessible through the openings explorer, this should be fairly easy to implement. What do you guys think? 

I'd had a similar thought. Thousands of games are played at each elo each day. Why not use those games as the basis for more "human" bots. That data for a 800 elo player would mean they make real errors instead of 100% perfect accuracy with intentional blunders thrown in randomly.